What is MongoDB? An Expert Guide - Oracle

- **Transaction support**. MongoDB transactional support is not as mature or robust as that found in traditional relational databases. Complex transactions, especially those that span multiple operations, may not perform as well and can be challenging to implement in MongoDB. - **Data consistency.** MongoDB’s use of “eventual consistency” for replica sets can lead to situations where all users aren’t reading the same data at the same time. For applications that demand strong consistency, this can be a serious drawback. - **Join operations.** MongoDB doesn’t support joins the way SQL databases do. It does, however, offer options that perform a similar function, though they are generally less efficient and can lead to more complex queries and slower performance—especially when dealing with complex relationships between documents. - **Memory use.** MongoDB stores its most frequently used data and indexes in RAM, so its performance is highly dependent on having sufficient RAM. As a result, a MongoDB database can consume more memory resources and, potentially, more hardware than other databases. - **Storage overhead.** The self-containing document paradigm used by MongoDB can lead to larger storage requirements compared to the highly normalized tables in relational databases. Additionally, MongoDB’s dynamic schema can cause data redundancy and fragmentation that can increase storage use—and costs. - **Indexing limitations.** MongoDB supports many indexing options, but maintaining a large number of indexes can degrade write performance. It’s just not built for frequent writes, because each write operation might need to update multiple indexes—often pitting query performance against write performance. - **Cost.** In scenarios where high availability and horizontal scaling are required, the cost associated with running and maintaining a MongoDB cluster—especially in cloud environments—can be significant. The need for lots of RAM and storage can also drive up costs. That’s especially true in high-availability situations where replica databases require an equal number of resources.